OK, I was able to reproduce the problem that Aaron had and figure out what's 
going on.  I
also realize from looking at the code that we figured all of this out once 
before a
couple of years ago. :-)

So, here's what's happening:

When you run 'ant freshStart junitAll', you also run 'deploySoap'.  There is a 
funky
interaction between deploySoap and JBlanket that causes deploySoap to bomb when 
jblanket
is enabled.  The build doesn't stop, in fact everything continues to work fine, 
coverage
data is collected and sent, etc.  It's just that the deploySoap target ends on 
a bad
note.

To fix this, you have to run deploySoap as a forked Java process.  Almost too
conveniently, there is a property called 'deploy.soap.fork' in the deploySoap 
task.  Set
this puppy to true, and 'ant freshStart junitAll' works fine with jblanket 
enabled.
Since this slows down the build, we keep it defaulted to false.  To override 
the default
setting and get deploySoap to play nice with JBlanket, just put the following 
in your
hackystat.properties file:

deploy.soap.fork=true

It also turns out that the Hackystat-ALL configuration (which runs JBlanket) is 
in fact
getting that same error in deploySoap; you just have to look pretty hard to 
find it since
the error is printed on one line.   Basically, we forgot to set 
deploy.soap.fork to true
in the hackystat.properties file for the hackystat-ALL configuration. I'm going 
to fix
that right now.

Cheers,
Philip

Reply via email to